🌿 Leaf Health Classification (Healthy vs. Diseased Linden Tree Leaves)

This project provides a machine learning model to classify linden tree leaves as either healthy or ill (diseased).
The model was trained and optimized using Edge Impulse and exported as an Arduino library.

The resulting model can run directly on Arduino boards that support TensorFlow Lite Micro.

---

🚀 Features

• Pre-trained Edge Impulse model for Arduino-compatible MCUs
• Classifies healthy vs. ill linden leaves in real-time
• Packaged as a ready-to-use Arduino library

---

🔧 Installation Tutorial: Deploying the Model on Arduino

Follow these steps to use the model on your Arduino board:

1. Install Arduino IDE

• Download and install the Arduino IDE.

2. Install Required Libraries

• Open Arduino IDE → Sketch → Include Library → Manage Libraries…
• Install:
  • Arduino_TensorFlowLite (official TensorFlow Lite Micro support)
  • Arduino_LSM9DS1 or other sensor libraries (if required by your board)

3. Add the Edge Impulse Model Library

• Download mobile_160_035_8_05_97Test.zip from this repository.
• In Arduino IDE, go to Sketch → Include Library → Add .ZIP Library…
• Select the ZIP file.
• The model will now appear under:
  File → Examples → mobile_160_035_8_05_97Test

4. Run Example Code

• Open: File → Examples → mobile_160_035_8_05_97Test → esp32 → esp32_camera
• Select your board (e.g., ESP32S3 Dev Module)
• Upload the sketch

5. Test the Model

• Provide input data (image or sensor pipeline as used in Edge Impulse).
• The model outputs:
  • 0 → Healthy Leaf
  • 1 → Diseased Leaf

---

🧪 Dataset

• Images of healthy and diseased linden tree leaves
• Preprocessed into input size 160×160 during Edge Impulse training

---

⚡ Model Details

• Framework: Edge Impulse (TensorFlow Lite Micro)
• Architecture: MobileNet (Tiny variant)
• Input: 160x160x3 RGB image
• Output: Binary classification (Healthy / Ill)
• Deployment: Optimized for low-power MCUs

---

📖 References

• Edge Impulse Docs
• TensorFlow Lite for Microcontrollers
• Arduino + Edge Impulse Guide

---

📌 Future Work

• Improve accuracy with more training data
• Extend classification to multiple tree species
• Add on-board image capture + preprocessing pipeline